All case studies
Composite manufacturing

X-ray defect detection on carbon-fibre structural parts

A composite-materials manufacturer put computer vision on its X-ray line to catch delamination earlier, steady quality across shifts, and keep expert judgement in the loop.

98%
detection accuracy on graded parts
40%
less unplanned line downtime
50%
fewer escaped defects
8 wk
data to first working POC
DELAMINATIONFLAGGED · NGRISK QUEUErank 1 · NGVision ranks the riskiest image first; people confirm.
Background

The manufacturer produces carbon-fibre structural parts for performance brands. Every part is X-rayed and a trained inspector decides pass or fail. As volume grew across two factories and several X-ray machines, that manual step became the bottleneck and the source of the most expensive escapes.

Judgement varied by inspector and by shift, the hardest defect (delamination) was easy to miss, and images came from machines with different characteristics, so no single rule of thumb held across the line.

Business challenges

Manual review could not keep up

Every scan waited for a person, and throughput dropped whenever an experienced inspector was off.

The costly defect was the easiest to miss

Delamination is subtle on an X-ray, and it is exactly the defect customers care about most.

Several machines, different images

An older CT and newer scanners produced different image profiles, so a model tuned on one did not transfer.

No labelled OK / NG history

There was no clean dataset to train a classic supervised model from day one.

Solution

We started unsupervised. An anomaly model ranks every image by how unusual it looks, rolled up to the part by barcode, so inspectors review the riskiest images first instead of all of them. Each machine gets its own calibration, and same-machine comparisons are kept separate from cross-machine ones.

A review workstation lets inspectors confirm or overturn the ranking, and those labels feed a growing hard-case library. The system never auto-promotes a model or hides uncertainty: when it does not have the data, it says so.

Scope
Duration

8 weeks, data to first working POC

Rapid design & deployment

One expensive problem, proven before it scales. The A1 to POC method, run in weeks not quarters.

A1 · Confirm the goal
Pin the one defect and the pass mark, agreed with QE before any model.
A2 · Audit the data
Map machines, image profiles, and what labels actually exist.
A3 · Design the use case
Choose ranking over classification, and decide who reviews what.
POC · Build & validate
Train, validate against held-out parts, tune per machine.
Handoff · Review workstation
Ship the workstation, the label flow, and a named owner.
Before & after
Before
After
Inspection
Every scan reviewed by hand
Vision ranks images, people confirm the top
Escaped defects
Delamination slipped through on busy shifts
Halved, with the hard case caught earlier
Consistency
Varied by inspector and machine
Per-machine calibrated, steady across shifts
Line downtime
Inspection backlog stalled the line
40% less unplanned downtime
Return on investment

The payback came from escapes avoided, not headcount cut.

Governance & standards

We design and deploy to the ISO/IEC 27001 (information security) and ISO/IEC 42001 (AI management system) frameworks. Data stays where it should, decisions that carry real cost keep a human in the loop, and every model call is logged for audit.

ISO/IEC 27001
Information security management
ISO/IEC 42001
AI management system

Designed and deployed to these frameworks. Not a certification claim.

Related use cases
Contact

Have a problem like this?

Bring your messiest operational problem. In 30 minutes we will tell you where AI would pay back fastest, and where it would not.

Discuss an operations problem